AI spending boom to cool as enterprises hit pause until 2027

Analysts predict deflation as gap between promises and reality widens

Analysts have warned that the global AI market may be heading for its first major correction, as large enterprises push back a significant portion of their planned spending.

According to new research from Forrester, 25% of enterprise AI investments slated for next year will be deferred until 2027, signalling a slowdown in deployment and a cooling of the AI boom.

Forrester attributes the looming slowdown to a growing disconnect between vendor promises and actual business results.

While AI providers continue to tout revolutionary productivity gains, many organisations are struggling to translate pilot projects into measurable financial returns.

The research found fewer than a third of corporate decision-makers could directly link AI initiatives to tangible financial growth.

Forrester warns that AI investments will increasingly be judged by their returns next year. Financial rigour will slow production deployments, pushing about 25% of planned AI spending to 2027.

The report predicts this will expose weaknesses across the AI supply chain, driving a shakeout among vendors and forcing a market correction.

The company says AI will be used less efficiently as demand slips, so the best quality results will remain expensive to generate. Providers will offer significant discounts to attract new customers.

“Savvy buyers” should take advantage of this weakness to cut their AI costs and re-focus their spending on "top- and bottom-line impact” - that is, projects that actually generate revenue.

Bubble or correction

Forrester's forecast contrasts with a more optimistic outlook from rival research firm Gartner, which projects worldwide spending on AI to reach nearly $1.5 trillion in 2025.

Gartner's Distinguished VP Analyst John-David Lovelock acknowledged that the sector faces consolidation, but dismissed the idea of a bursting bubble.

The AI market will, he admitted, experience an “extinction event” among gen AI model providers, referring to smaller vendors unable to meet the escalating compute demands of large models.

But, he said, that is "not the indication of a bubble.” Instead, it's a reallocation: mergers, divestitures and other investments will keep money in the market.

Nonetheless, Gartner's own AI projections have grown more cautious.

Earlier this year, the firm predicted that 40% of agentic AI projects – autonomous systems designed to execute complex multi-step tasks – will be cancelled by 2027.

Supporting this scepticism, a Carnegie Mellon University study in mid-2024 found that AI agents succeed only 30 to 35% of the time on multi-step operations.

And just last month, OpenAI co-founder Andrej Karpathy said that true AI agents are still a decade away.

Beyond the tech industry, financial institutions are also raising red flags.

The Bank of England recently warned of a potential "sudden correction" in equity markets driven by inflated AI valuations.

The International Monetary Fund expressed similar concerns in its World Economic Outlook report last month, describing AI investment trends as reminiscent of the internet frenzy that preceded the 2000 crash.

Despite growing caution among investors and policymakers, industry leaders continue to double down on AI infrastructure.

OpenAI, widely viewed as the centre of the AI revolution, has announced a series of massive deals in recent months.

Last month the firm revealed a hardware partnership with Broadcom to co-develop and deploy 10 gigawatts of custom AI accelerator hardware between 2026 and 2029. It also signed a new deal with Microsoft, which now owns more than a quarter of the ChatGPT developer.

The Broadcom agreement follows OpenAI's six-gigawatt chip purchase from AMD, valued in the tens of billions of dollars, announced in September.

That same month, Nvidia disclosed a $100 billion investment in OpenAI and a letter of intent to supply an additional 10 gigawatts of AI compute capacity.